Oct 16, 2026Added Oct 19
Huntress reports that a compromised ConnectWise Cloud Backup (formerly SkyKick) app credential was used from 144.172.64[.]128 to obtain Microsoft Graph tokens in customer M365 tenants, bypassing user MFA. Unauthorized sign-ins were confirmed in 162 tenants as of Oct 16; Huntress later saw almost 1,000 Entra tenants with sign-ins from the IP (300+ successful). ConnectWise disabled relevant apps as of Oct 17. Huntress's Oct 19 (1:30 PM ET) update: across 10,000+ tenants with ConnectWise apps, 144.172.64[.]128 is the only public IP that produced application sign-ins; 144.172.64[.]99 was a concentrated user password spray (Oct 10-14), 144.172.64[.]60 a few failed logins (Oct 18), and 144.172.94[.]243 an ongoing broad failed-login spray, none currently tied to the app sign-ins. A second tenant with post-compromise activity was found: in both, the app granted itself Global Administrator and planted secrets on a few in-tenant apps around 03:45 UTC Oct 15. Last confirmed malicious sign-in was 08:44 UTC Oct 15. No mailbox, file, user, or policy changes were seen, but Graph directory reads leave no audit record. Disabling apps does not remove credentials already added.
Source · huntress.com
Oct 19, 2026Added Oct 20
Palo Alto Networks launched Cortex XDR 6.0 and Cortex XDR One on Oct 19. XDR One combines Cortex XDR and Agentic Endpoint Security (AES, from Koi) in a single sensor and console. It discovers and risk-scores every agent, MCP server, plugin, model, and extension. AI Detection and Response (AIDR) records each AI coding agent session (prompts, shell commands, file reads, tool calls, network calls) and blocks policy-violating actions before they execute, covering tools such as Cursor, Claude Code, GitHub Copilot CLI, Codex CLI, Antigravity CLI, and Kiro. Also new: Native App Control, the Cortex Targeted Sensor (AES or AD-SPM alongside a third-party EDR), on-endpoint small language models, extended on-write WildFire checks, and AI-powered YARA rule management via AgentiX. Palo Alto says XDR One cuts false positives by up to 95% and gives Cortex XSIAM the data to govern AI risk.
Source · paloaltonetworks.com
Oct 19, 2026Added Oct 20
Announced Oct 19, Unit 42 Managed AES adds 24/7 Unit 42 expertise on top of Cortex AES across four areas: AI governance, threat hunting, investigation and response, and AI-related cyber hygiene (AI Security Posture Reports and Threat Impact Reports). It works with standalone Cortex AES alongside an existing EDR or with Cortex XDR One, and Palo Alto says Unit 42 MDR will include it. Palo Alto cites that only 6% of organizations have an advanced AI security strategy.
Source · paloaltonetworks.com
Oct 20, 2026Added Oct 21
SDxCentral (Oct 20) reports that Cortex XDR 6.0 intercepts AI prompts and responses at the operating-system level, so protection is not limited to apps with native monitoring hooks; named coverage includes Claude Code, Codex CLI, and open-source frameworks such as OpenClaw. The release adds preconfigured AI runtime security policies mapped to MITRE ATLAS, small language models on the endpoint sensor that flag malicious or obfuscated scripts before execution without a cloud round trip, AgentiX-assisted YARA rule creation and fleet sweeps (matches return hashes, signing info, and host details), and native App Control on Windows and macOS. XDR One combines XDR with Koi's Agentic Endpoint Security in one sensor and console, and Cortex Targeted Sensor runs selected capabilities next to a third-party EDR.
Source · sdxcentral.com
Oct 21, 2026Added Oct 22
In its Oct 21 NVIDIA OpenShell blog, Palo Alto says it has announced an integration of Cortex XSIAM with NVIDIA DOCA Argus. DOCA Argus monitors host activity from BlueField's independent trust domain, so XSIAM gets runtime signals that don't depend on an agent on the monitored host. NVIDIA Sentry, built on DOCA Argus, inspects memory and network activity out-of-band and streams alerts straight to XSIAM, which correlates them with network, cloud, and identity data. Palo Alto says it is continuing to develop broader fleet visibility and response workflows.
Source · paloaltonetworks.com
Oct 21, 2026Added Oct 24
CORRECTION (Oct 25): Unit 42 retracted its Oct 23 update that tied a European government agency compromise, with agentic-AI post-exploitation reaching password hashes and MFA secrets within two hours, to PHP2Shell: 'We are unable to confirm that initial access is associated with the specific PHP2Shell CVEs.' The intrusion was real; only the CVE link was withdrawn. Still valid: Cortex Xpanse counts over 22.6 million potentially vulnerable exposed instances (CVSS 4.0 score 9.2), and Unit 42 Managed Threat Hunting published two XQL queries: post-exploitation commands under PHP web workers (xdr_data) and inbound exploit attempts in NGFW threat logs (panw_ngfw_threat_raw, TID 543004). Fix: patch PHP to 8.5.12, 8.4.27, 8.3.36, or 8.2.35, rebuild images, restart workers, and enable Threat IDs 543004 and 97622-97625.
Source · unit42.paloaltonetworks.com
Oct 22, 2026Added Oct 23
Palo Alto's Oct 22 Cortex Cloud blog covers CVE-2026-109618 and CVE-2026-109619 layer by layer: Cortex CLI blocks vulnerable images in CI/CD, Cortex Cloud Vulnerability Management finds vulnerable PHP, Cortex ASM has a PHP Attack Surface Rule, XDR and the Network Scanner inventory PHP on servers, Cortex Cloud Runtime and XSIAM handle post-exploitation, and NGFW with ATP Plus blocks exploit attempts (TID 543004). Affected deployments include PHP-FPM, Apache mod_php, php-cgi, FrankenPHP, and LiteSpeed. The blog also introduces a custom Cortex AgentiX agent, the PHP2Shell Security Advisor, which finds affected VMs, images, containers, and serverless functions, ranks them by criticality, checks XDR agent coverage, and recommends fixes. It runs read-only; remediation stays with the team.
Source · paloaltonetworks.com
URGENTOct 21, 2026Added Oct 23
CISA's Oct 21 alert tells organizations to patch PHP immediately; the NSA issued a matching advisory the same day. Fixed PHP versions are 8.5.12, 8.4.27, 8.3.36, and 8.2.35 (or a verified vendor backport). CISA notes PHP is often embedded in other software. Systems that can't be patched now, or run PHP 8.1 or older, should be taken offline until fixed. CISA says a WAF may provide some protection, and recommends hunting for PHP worker processes launching shells or downloading tools.
Source · cisa.gov
Oct 22, 2026Added Oct 23
Unit 42's Oct 22 report says attackers increasingly steal OAuth access and refresh tokens instead of passwords; stolen tokens bypass MFA, survive password resets, and work until explicitly revoked. Campaigns fall into account hijacking (device-code phishing, authorization-code theft via legitimate apps' redirect_uri values), permission hijacking (consent phishing with vague or impersonating app names), and malicious redirects (abusing OAuth error flows). Recommended controls: disable end-user consent with admin approval, block unverified apps, enforce least-privilege scopes, revoke risky grants, audit redirect_uri values, and train users not to paste device codes.
Source · unit42.paloaltonetworks.com
Oct 22, 2026Added Oct 23
SecurityBrief's Oct 22 article on Cortex XDR One quotes Orcun Tezel (VP Technical Solutions, APJ). It says XDR One uses small language models on the sensor to block fileless attacks offline, claims up to 95% fewer false positives, and notes Agentic Endpoint Security can run standalone alongside other EDRs.
Source · securitybrief.news
Oct 23, 2026Added Oct 24
PHP disclosed CVE-2026-115560 on Oct 23, a session upload-progress RCE that only works when session.serialize_handler is php_binary or php_serialize and is switched to php via ini_set() before session_start(), with upload progress enabled. It will be fixed in the next regular release, not an emergency one. Mitigation: set session.upload_progress.enabled = 0 and restart PHP workers. CERT-EU says PoCs are circulating for the PHP2Shell family and advises patching to 8.5.12, 8.4.27, 8.3.36, or 8.2.35 or isolating systems, and checking the NCSC-NL list for products that embed PHP.
Source · theregister.com
URGENTOct 24, 2026Added Oct 25New
TechCrunch (Oct 24) reports that as of late Friday there are confirmed compromises and post-exploit activity, per Palo Alto Networks and Trail of Bits. Amazon Threat Intelligence saw a PHP2Shell exploit attempt in its MadPot honeypots on Oct 17, before disclosure, and advises looking for evidence from that date (POST requests with PHP_SESSION_UPLOAD_PROGRESS, unexpected process execution). VulnCheck built an exploit that bypasses ASLR on x86-64 with PHP's recommended production settings, and SOCRadar found 'one-click' exploit tooling for PHP 8.4.26 under Apache mod_php. Fix: patch to PHP 8.5.12, 8.4.27, 8.3.36, or 8.2.35 and restart workers; isolate EOL PHP; rotate secrets on exposed unpatched hosts; run the Unit 42 XQL hunts back to Oct 17.
Source · techcrunch.com